Green Data Part 1: Optimizing Solar Production with Data Science and IoT

Contributed by

7 min read

I know that I am the dangerous type of individual. I am that guy who knows just enough about many subjects to be slightly scary to individuals who truly understand a given discipline. One of those areas is Solar Power. After my first array was installed, I became obsessed with understanding solar power. This has led me to this year, when I installed my third array, a ground-mounted array with a single axis solar tracker from U.S. Solar Mounts (https://www.ussolarmounts.us/).

After getting it up and running, I realized that the optical tracking of the panels was not ideal. I knew I had a problem when I witnessed, a number of times, one of the physical arrays pointing east at the same time the other pointed west,. As I thought through said problem, I theorized that I could do better with IoT, data science, and a little bit of math, all areas that I know just enough to be dangerous. So here is my journey through a series of five initial blogs (including this introduction). In post 2, we will discuss the hardware I used. Posts 3 and 4 will be data analysis and then calibration and putting into operation a more optimal solution. Finally, in post 5, we will review the results.

Background: Solar Arrays

Let's walk through a bit of the history of the arrays that I have to this point.

Array 1: Professionally Installed, Fixed-Tilt, Roof-Mounted Array

In 2017, I had my first solar array installed: a roof-mounted solar array that operates at 6.6 kWp on a 6 kW SolarEdge inverter. For those who haven't dug into the solar industry, the kWp is essentially the maximum rating of each panel times the number of panels. For my roof mount system, each panel produces 300 watts, and I have 22 panels; therefore, my kWp is 300W * 22 panels or 6,600 watts (6.6 kWp). This isn't what it produces, but the max it is capable of producing if all variables, like temperature and solar angle, are absolutely perfect.

This rooftop array was installed by my friends at North Wind Renewable Energy (http://www.northwindre.com). They did a fantastic job and even allowed me to ask hundreds of questions during the install. This was good, because during the process I became obsessed with how it worked, how it integrated with the electrical grid, and by the end of the install, how I could add more.

Array 2: Ground-Mounted Array – Manually Adjustable Tilt

After that summer install, I went into overdrive, researching if I could install an array myself. The answer? Yes! That fall, I installed a second array. This one is a ground mount unit on which I could manually adjust the angle of the panels a few times a year to optimize the collection. It is a 7.14 kWp array on a 7.6 kW SolarEdge inverter. This was a very satisfying project and allowed me to compare an array on a fixed plane (the roof-mounted array) with an array on an adjustable mount.

I enjoyed having the two arrays because it allowed me as a data scientist to collect data from both and compare. I did many comparisons and would like to do a follow-up post that, once I've collected data from all parts of the year, will compare the three types of arrays I have.

Array 3: Ground-Mounted Array with Single Axis Tracking

My newest array, and the array we'll be working with in this blog post, is actually two separate arrays as you can see from the picture:

The array by the barn-shaped house is my second array, and the two arrays on the right I consider to be one array because they connected through a single inverter. This third array consists of the two physical arrays. Each array has nine 305 W panels for a total of just under 5.5 kWp on a 6 kW SolarEdge inverter.

Defining the Problem: Optical Tracking

I thought that, ideally, optical tracking would be the best way to track the sun (especially since that's what the panels came with). Perhaps it's not adjusted correctly, or perhaps conditions where I am just don't favor optical tracking, but too many times I've looked at my arrays and seen them doing something similar to this:

It can't be right to have them pointed in completely different directions. There are a number of reasons for this, given that heavy cloud cover obscures and manipulates the "brightest" point the eye can see. Since the optical eye doesn't have a black backing like the panels, it doesn't generate much heat to melt snow or frost crystals, and just in general, the dark trees to the east and bright house to the west can manipulate the electronic eye to what it sees as bright.

IoT: Collecting and Controlling All the Things

In order to solve this problem, I needed data. Observing the panels out of whack every now and then let me theorize I had a problem, but how bad was it? How could I track when the array was non-optimal? How could I determine the optimal position? How could I control the array to put it where I wanted it to be? Lots of questions to be answered, but the first question before investing in a solution was to determine "could I identify the optimal position of the array?"

The answer to this question was yes! A paper by William F. Marion and Aron P. Dobos on the "Rotation Angle for the Optimum Tracking of One-Axis Trackers from the National Renewable Energy Laboratory" (https://www.nrel.gov/docs/fy13osti/58891.pdf) confirmed to me that it should be very possible to determine the optimum rotation angle on my single axis tracker.

I am going to be honest with my readers here: my math ability was not where it needed to be when I first read this paper.Reading through it over and over, my heart told me the answer was indeed in the paper, but my brain demanded coffee in order to find it. More on that later in the next blog.


This blog post was published December 05, 2018.
Categories

50,000+ of the smartest have already joined!

Stay ahead of the bleeding edge...get the best of Big Data in your inbox.


Get our latest posts in your inbox

Subscribe Now